The ease of programmability in Software-Defined Networking has greatly facilitated SDN application development and deployment. Applications run simultaneously in the controller and generate policies to serve the network together. However, multiple applications may cause unintentionally harmful interferences even if each SDN application is properly programmed. Unfortunately, the existing SDN verification and test work have no consideration of application interferences. To solve this problem, we develop the mathematical models for applications and policies and elaborate the relationships of policies based on computational geometry. The application interference is formally defined according to policy conflict. To this end, we propose an efficient SDN Application Interference Detection and Elimination (SAIDE) scheme. Firstly, on the basis of designed policy refactor, we encode the matching fields of policies as bit vectors and design bit vectors AND operation to analyze the policy relationships. Then, we combine the action and refactor fields of policies to detect the application interferences (direct and indirect interferences). Finally, with multi-criteria decision making, we assign priorities to conflicting policies to eliminate the corresponding application interferences. We demonstrate the effectiveness and scalability of SAIDE through a proof-of-concept prototype. Simulation results show that the SAIDE can effectively detect the application interferences with 98% accuracy at least and has reduced the detection time by 23.9% compared to the state-of-the-art work.

软件定义网络中易于编程的功能极大地促进了SDN应用程序的开发和部署。应用程序在控制器中同时运行，并生成策略以共同为网络服务。但是，即使正确编程了每个SDN应用程序，多个应用程序也可能造成意外的有害干扰。不幸的是，现有的SDN验证和测试工作没有考虑应用程序干扰。为了解决这个问题，我们为应用程序和策略开发了数学模型，并根据计算几何学阐述了策略之间的关系。根据策略冲突来正式定义应用程序干扰。为此，我们提出了一种有效的S DN A应用I到的干扰d etection和ë limination（SAIDE）方案。首先，在设计策略重构的基础上，将策略的匹配字段编码为比特向量，并设计比特向量AND运算，以分析策略关系。然后，我们结合策略的操作和重构字段来检测应用程序干扰（直接和间接干扰）。最后，通过多标准决策，我们为冲突策略分配了优先级，以消除相应的应用程序干扰。我们通过概念验证原型展示了SAIDE的有效性和可扩展性。仿真结果表明，与最新技术相比，SAIDE可以至少以98％的准确度有效检测应用干扰，并且将检测时间减少了23.9％。